SLOPSHOPPER

eta

Shows long-running and background jobs with a percentage and an estimated finish time, read live from their output.

newbandrowsguardstatusprompt
v0.2.2MITupdated 2026-10-16HarzerHeribert/claude-modpack/eta
A shopper browsing a rack in a slop shop
README

<img alt="claude-modpack: eight Claude Code mods, each measured on real tasks before it shipped, built to work together. statusline, smart-compact, delegate, lsp-plus, cells, eta, seek, route." src="assets/banner-light.svg" width="880">

<h3>Install in Claude Code</h3>

/plugin install modpack --marketplace HarzerHeribert/claude-modpack

Answer y, then run /modpack:setup to set everything up step by step.<br> Prefer doing it by hand? SETUP.md has every step.

Release License: MIT Claude Code 2.1.296 730 graded runs, all correct 8 mods Lab notes

Claude Code with fewer wasted steps, and a clear view of what it's doing.

Eight free add-ons that work on their own. One command installs them, and there is nothing new to learn.<br> Every number on this page was measured, including where it doesn't help.

Install · What you get · Honest numbers · The site: ideas and evidence · Lab notes · Questions

<img alt="Animated: the cost per job with and without the pack, as bars that grow: in a long session $0.44 against $0.29 (−33%; measured twice, −26% one day and −40% the next), bigger jobs in a fresh session $0.55 against $0.51 (−8%), quick questions in a fresh session $0.23 against $0.24 (+2%, about even: +5% one day, −4% the next); then four numbers: −33% cost per job in a long session, −21% steps for the same work, −22% cost of a feature a cheaper model builds from your plan, 224 of 224 results correct." src="assets/numbers.gif" width="880">

Why it saves

Every time Claude takes a step (reads a file, makes an edit, runs the tests), it reads your whole conversation again. That re-reading is most of what a session costs, and it grows the longer you work. So the cheapest step is the one Claude never has to take.

The pack takes out steps that need no thinking. Several edits and their test run as one step. A search for a name in your code comes back with the exact answer. And when you approve a plan, a cheaper model does the building while Claude checks the result. The work comes out the same; Claude just gets there in fewer round trips.

What you get

Always know where your plan stands

<img alt="A real Claude Code session: Claude starts the stress tests in the background; above the prompt a row shows them with a bar, a percentage and a finish time, climbing from 8% to 100%; under the prompt the status line shows the model, the context, the plan's 5-hour and weekly use and a green ready dot." src="assets/demo-band.gif" width="880">

One line under the prompt: the model, how full the conversation is, and how much of your plan's 5-hour and weekly limits you have used, with when each one resets. Long jobs get a progress bar and a finish time above the prompt. And when you step away, your session is kept ready, or tidied up before it goes cold, whichever costs less; the line tells you which.

These are the statusline, eta and smart-compact mods.

Several changes in one step, and you can still undo them

<img alt="A real Claude Code session: asked to add two docstrings and run the whole test suite, Claude runs one cell; an orange card shows the read, the two edits, the diff and the test run live, and the tests pass: 1990 passed." src="assets/demo-cells.gif" width="880">

Claude makes several edits and runs the tests as one step, and you watch each part happen in a live card. Every edit still goes through Claude Code's own edit tool: you see each change, and /rewind undoes it.

That's the cells mod.

A cheaper model builds what you approved

<img alt="A real Claude Code session: in plan mode Claude plans a new click.Duration parameter type; the plan is approved; Claude hands it to a Sonnet worker, whose orange card shows its steps live with a row above the prompt (hand-off 1, sonnet medium, requests and cost); then Claude checks the diff and reports: the new type works and the full suite passes." src="assets/demo-delegate.gif" width="880">

When you approve a plan, Claude hands the building to Sonnet, a cheaper model, and checks its work before reporting back. In our tests that made a feature 22% cheaper and new tests 26% cheaper, with every result correct (corrected 16 Oct: worker costs were undercounted; it was 27% and 33%). You can watch the helper work in its card, and /delegate shows every step it took.

That's the delegate mod.

Feature work goes to the cheaper model by itself (opt-in)

A small, fast classifier reads each prompt you send. Only when it is feature or test work is Claude told to hand it straight to delegate's worker: the worker reads the code and builds, delegate checks the result itself, and Claude reports after two steps of its own. On a TypeScript feature that made the job 51% cheaper and 42% faster than the pack alone, with every result correct (corrected 16 Oct: worker costs were undercounted; it was 57%); everything else stays with Claude. The classifier is an outside service (TypeSafe's Jev or ScaleDown), so this is off until you turn it on and give it a key: from then on, the start of each prompt goes to it, in the US, to be classified.

That's the route mod.

Straight answers about your code

When Claude searches your code for a name, it gets the exact list of places that use it, from the same engine your editor uses, instead of reading through search results. Renames and "who calls this?" questions take fewer steps. New in 0.5.0: Go, Kotlin and Ruby code, shell scripts and Terraform. Run /lsp-plus:setup once in your project to turn it on.

That's the lsp-plus mod.

Fewer, leaner steps when Claude looks around

Before Claude changes anything, it looks around your code, and everything it reads stays in the conversation, read again with every later step. Beside the first prompt you send in a session, and again as the session grows, the pack passes on your preference to look around in few, lean steps: the searches it can already name in one go, the lines that matter rather than whole files, nothing read twice. Against the pack without it, that made jobs 12% cheaper in a fresh session, 16% on a TypeScript library and 10 to 16% in long sessions, at the same quality. It runs on your machine and sends nothing anywhere.

That's the seek mod.

You stay in control

<img alt="A real Claude Code session: /mods lists the pack's mods (recorded with the six of 0.4.0, before route and seek) and opens a pane of cards; the status line is turned off and disappears from under the prompt, then turned on again." src="assets/demo-mods.gif" width="880">

/mods lists the eight and turns any of them on or off, right away, with a click or by typing, such as /mods statusline off. When something can't fully work on your computer, the pack tells you once, with the command that fixes it. When another plugin's permission or notification hook hangs (Warp's does outside the Warp terminal), the pack stops waiting for it after 10 seconds, so Claude Code can't freeze, and tells you which one it was. It keeps itself up to date (or only lets you know, if you prefer).

That's the modpack plugin, which installs the other eight.

Install

In Claude Code, type:

/plugin install modpack --marketplace HarzerHeribert/claude-modpack

Answer y to add the marketplace, then pick where to install it. That installs all eight. Then, in the project you work in:

/modpack:setup

It checks your machine, asks once what you want on (route stays off unless you say yes), sets each mod up in order, including the language servers for the languages your project uses, and checks that each one works. Done. /mods shows what's on.

By hand instead: SETUP.md has every step.

Each mod installs on its own the same way, for example /plugin install smart-compact --marketplace HarzerHeribert/claude-modpack. lsp-plus, delegate and cells build on Claude Code's LSP tool, which needs a language server for each language: Language servers lists them.

Honest numbers

We measured the pack on Claude Opus 5.5 doing seven real jobs in an open-source project (click), from a one-line lookup to building a feature, on two days, 13 and 15 October. Every result was checked by hidden tests. The same jobs ran with the pack and without it, and we compared what they cost at API prices.

Cost per jobWithoutWith the pack
In a long session (Claude has already read the whole project)$0.44$0.29−33%
Bigger jobs in a long session (a feature, tests, a rename)$0.91$0.57−37%
In a fresh session$0.37$0.35−4%
Bigger jobs in a fresh session$0.55$0.51−8%, and 15% faster
Quick questions in a fresh session$0.23$0.24+2%, about even
  • Where it doesn't help: quick questions in a brand-new session. The pack's tools come along with every step, and a three-step lookup has little to save: 5% more one day, 4% less the next, about even.
  • Writing tests: the pack alone saves nothing there (−2%, 12 runs each way): every step waits on the last test run. route hands it to the worker instead (−36% to −46%, below).
  • Nothing got worse: all 224 runs behind this table produced a correct result, with the pack and without.
  • The 95% ranges: long session −39% to −27%; fresh −7% to −1%; bigger jobs fresh −12% to −3%; quick questions −1% to +6%. The long session is two runs per job each day, and the days differ: −26% on one, −40% on the other, mostly from the feature (plain needed 12 requests one day, 16 the next).
  • The same on other code: on a TypeScript library (chrono) the pack cost 7.6% less (−12% to −3%) and took fewer steps on all eight jobs. On 30 real bug reports from SWE-bench Verified it fixed 30 against 29, with 23% fewer steps; the cost, −12%, is not certain at one attempt per bug (−33% to +12%).
  • A plan handed to a cheaper model: −22% on a feature, −26% on tests, against Opus doing it alone with the pack (16 runs, all correct).
  • With route turned on: a further −14% on click's eight jobs (route 0.1). With 0.6.0's verified hand-offs, against the pack in the same batch on TypeScript: a feature −51% and 42% faster, writing tests −36%; Claude itself takes 2 steps instead of 6 to 11 (24 runs, all correct).
  • Corrected 16 Oct: worker costs were undercounted. A worker's transcript records its output as each response began, and our bench counted it from there, so every number with a hand-off in it came out too good. Corrected from Claude Code's own totals: a planned feature −22% (was −27%), tests −26% (was −33%), route on click −14% (was −17%), the TypeScript feature −51% (was −57%). Every saving is still a saving, a smaller one. The table above has no hand-offs and did not change. Each claim, old and new: CORRECTION-1016.md.
  • With seek (new in 0.6.0, measured against the pack rather than plain Claude Code, so not in the table above): −12% fresh on click, −16% on TypeScript, −10% to −16% in long sessions (seek, below).
  • Two days, two versions: the 13 October runs used the pack as of 0.3.0 and Claude Code 2.1.296, the 15 October runs 0.4.0 and 2.1.299. What changed in the pack between them (on/off switches, setup, messages shown when a worker is refused) never reached the model in these runs: its first request was 26.7k tokens plain and 30.6k with the pack on both days. The second day's plain runs were done twice, because the first try had the pack installed by mistake: lab notes 8 tells how we caught it.

<img alt="Cost per task, plain Claude Code against the modpack, in a fresh session and in a session already holding 216k tokens." src="assets/cost-light.svg" width="880">

<img alt="Model requests per task, plain Claude Code against the modpack: 3.2 against 5.6 in a session already holding 216k tokens." src="assets/requests-light.svg" width="880">

<img alt="Cost per task, Opus alone against a plan handed to a worker: feature work 22% less, tests 26% less." src="assets/delegate-light.svg" width="880">

How we measured, every run's record, and the scripts to run it all again are in benchmarks/, starting with SUITE.md. What we tried that did not make it, a 10,000-star code graph among it, is in the lab notes (as Markdown). Not measured: what smart-compact saves you while you're away (worked out from prices, not tested) and Windows.

Questions

Will Claude's work get worse? No. In every run we graded, 730 in all, the result was correct with the pack and without, and every one was graded again for this release. The pack changes how many steps Claude takes to get there, and in our tests the results came out the same.

Is it safe? It's open source (MIT), so every line is here to read. Code that Claude runs inside a cell runs in a sandbox with no network access. Edits go through Claude Code's own edit tool, so your usual permission prompts and /rewind still apply. The pack sends nothing anywhere except a once-a-day check on GitHub for a new release, which you can turn off, and, only if you turn route on, the start of each prompt to the classifier you chose (TypeSafe or ScaleDown).

Does the pack add anything to my prompts? Your prompt itself is never changed. Beside it, Claude can get a short note in your name: seek's setting to look around in few, lean steps (with the first prompt of a session, and again after a compaction), and, only if you turn route on, the order to hand feature or test work to the worker; delegate gives the same order when you approve a plan of 20 lines or more. A subagent Claude starts gets cells' and seek's settings at the end of its task, since your prompts never reach it. /mods seek off and /mods route off stop theirs; subagents: false in cells or seek stops theirs for subagents.

I'm on a Pro or Max plan. Does this help me? You don't pay per job, but every step counts against your 5-hour and weekly limits. We measured cost at API prices, not plan usage, so we can't promise a number. Fewer, lighter steps should leave more of your limits, and the status line shows exactly where you stand.

Does it work in the desktop app? Yes, cards, progress and the status line included. One limit: in auto mode the desktop app doesn't let an add-on start a helper model, so plans aren't handed off there (turn route off there too); in manual mode, or in the terminal, they are. Windows hasn't been tested yet.

How do I turn it off? /mods switches each part off or on, right away. /plugin removes them.

Were any of these numbers wrong? Yes, the hand-off ones. On 16 October we found that our bench had counted a worker's output only as each of its responses began, so every saving that involved a worker came out too large: a planned feature was −27% and is −22%. They are corrected here, the old and new number side by side in CORRECTION-1016.md. The pack-against-plain numbers had no workers and stand.

What does it cost? Nothing. It's free and open source.

The mods in depth

One line under the prompt, drawn in Claude Code's own hint row so the mode indicator and the status share it:

⏵⏵ auto mode on       Opus 5.5 xhigh   412k ━━━━━━──────   5h 34% ↺2h14   7d 12% ↺3d   ● ready · cold in 44m
  • Its own segments: the model and its effort, the context against the auto-compact point (a highlight sweeps along it while Claude works), and how much of the plan's 5-hour and 7-day windows is used, with when each resets: quiet under half, plain past it, amber from 80%.
  • Other mods' segments: a mod calls $.statusline.set({ id, priority, parts, short?, side? }) and lists statusline among its dependencies. A part is text, a countdown ({ until, prefix?, warnMs? }, counted by the line) or a spinner ({ spin: true }), so a mod sends a segment when it changes, not every second. $.statusline.remove({ id }) takes it off.
  • Narrow terminals drop the least important first: the model, then the windows' resets, the windows, the gauge, and another mod's countdown before its own words.
  • Published pages: a pill on the left lists the pages this session published with the Artifact tool.
  • Phone and IDE: where no line is drawn, the same line is the session's status text, in a session the phone or IDE hosts. A terminal session watched over Remote Control (/rc) from the desktop app, the web or the phone gets no mod drawing and no status text as of Claude Code 2.1.296.
OptionDefault
enabledtruefalse turns the mod off; /mods does it for all of the pack.
placementlinetail: appended dim to Claude Code's hint. band: above the prompt. off.
limitsalwayshalf: a plan window only once it is half used. off: never.
animatetrueThe highlight along the gauge.

An idle session's prompt cache expires after its TTL: 1 hour on a subscription, 5 minutes in overage or on an API key. The next prompt then writes the whole context to the cache again, at 1.25x (5 min) or 2x (1 h) the input price. For a 400k-token Opus 5.5 session on the 1-hour cache:

OptionCost
Let it expire, re-cache on return$3.20
One keep-alive (re-read the cached prefix; a read resets the TTL)$0.09
Compact before it expires~$0.87, plus detail lost to the summary

A few minutes before the cache expires (2.5 min on the 1-hour TTL, 45 s on the 5-minute one), the smart strategy:

  1. Sends keep-alives while they cost less than compacting would, counting lost detail as compactPenaltyUsd. It stops after maxIdleHours.
  2. Then compacts while the cache is still warm.
  3. Leaves small sessions alone: below minSavingsUsd of re-cache cost, or below minCompactTokens of context, it never compacts.

It never compacts while the prompt holds a draft or agents are running; it sends a keep-alive to hold the cache instead. If the machine sleeps past the deadline, it marks the cache expired and does nothing. A failed action is retried twice, 20 s apart. The TTL is detected from the plan and corrected from the cache hits and misses it observes.

On statusline's line it says whether the session is ready to prompt or has gone cold: ● ready, · cold in m:ss (amber in the last five minutes), · compacts in … when an idle compaction is near, ○ cold, or what it is doing (compacting, keeping warm, paused). /cache shows the live figures and plan; /cache pause, /cache resume and /cache keepalive control it.

OptionDefault
enabledtruefalse turns the mod off; /mods does it for all of the pack.
strategysmartcompact: compact at the first expiry. keepalive: never compact. off: status line only.
cacheTtlauto1h or 5m to override detection.
minSavingsUsd0.25Below this re-cache cost, the cache is left to expire.
compactPenaltyUsd0.5Higher keeps the cache warm longer before compacting.
minCompactTokens100000Never compacts a smaller context.
maxIdleHours6Stops keep-alives after this long idle.
compactInstructionsemptyPassed to idle compactions, like /compact <text>.

Prices are list prices, so on a subscription they are API-equivalent estimates, not a bill.

Every request of the main model re-reads its whole context: at 600k tokens on Opus 5.5 that is about $0.12 a step before any work is done. A worker reads only its own small context. Claude calls hand_off with a task that stands alone, its kind (read: changes no file; mechanical: the task spells out every change; build: something is left to work out), the symbols or file:line places it touches (focus), and for edits a check command. The kind picks the worker, so Claude has no model to weigh: Haiku at low effort for read and mechanical work, Sonnet at medium for build work.

  1. Where it reaches. For each focus entry the mod asks the language server for references and callers (with lsp-plus, the docs that mention it come along) and tells the worker.
  2. The worker. A subagent, read-only for read, with Edit and Write for mechanical and build. The call waits for it inside the tool call (one request of the main model) and returns the report with the evidence; with background it returns at once and the report arrives as a message. A follow-up within 50 minutes goes to the same worker, its reading still in its context.
  3. The evidence. The diff against a git snapshot taken at the start, however the worker edited; the check command, run by the mod itself; references outside the changed files; what the work cost against the same steps on the main model.
  4. The review. The main model judges the diff and the check, not the report.
  5. The log. Each hand-off goes to ~/.claude/mods/delegate/log.jsonl; /delegate sums it up and opens a side pane. While a worker runs, its card is drawn like a cell (the task, a row per step, the state and cost), and with eta loaded it is a row above the prompt too.

Measured on 120 Opus runs, all correct: a feature or tests cost 23–35% less handed off; lookups, traces, explanations and small edits cost more. Opus does not hand off unless told to, so ask for it as an instruction: "hand the tests to a worker". An order the mod gives works as well as yours (8 of 8 runs, 22–26% less), so when you approve a plan of 20 lines or more, delegate gives that order: plan in plan mode, approve, and a Sonnet worker builds it. Corrected 16 Oct: worker costs were undercounted; these were 27–37% and 27–33%.

A verified hand-off (0.3.0). The worker ends its report with a VERIFY block: the command that runs only its new tests, and what it left open. d

Source 4 files
hooks/register.tsx 379 lines
1import { atom, read, update } from 'claude-code'
2import type { Color, EngineInterface, Register, Timer } from 'claude-code'
3
4import { compile, describe, feed } from './estimate'
5import { BEAT_MS, table } from './table'
6import type { Tone } from './table'
7import type { Job } from '../types'
8
9const jobs = atom({ plugin: 'eta', key: 'jobs' } as const, [] as Job[])
10const tzOffset = atom({ plugin: 'eta', key: 'tzOffsetMin' } as const, null as number | null)
11const seen = atom({ plugin: 'eta', key: 'seen' } as const, [] as string[])
12
13/** smart-compact's palette: in Claude Code's themes "inactive" is the readable grey, "subtle" the dim one. */
14const COLORS: Record<Tone, Color> = {
15  accent: 'claude',
16  good: 'success',
17  bad: 'error',
18  warn: 'warning',
19  text: 'text',
20  label: 'inactive',
21  faint: 'subtle',
22  glint: 'text',
23}
24
25const TOOL = 'track_progress'
26const POLL_MS = 4000
27/** A failure the person has not seen stays this long at most. */
28const KEEP_FAILED_MS = 10 * 60_000
29/** A job that writes its state and whose file has not grown this long has stopped. */
30const STALE_MS = 20 * 60_000
31const CHUNK = 512 * 1024
32const CHUNKS_PER_TICK = 32
33const REMOTE = new Set(['mobile', 'vscode'])
34
35const TOOL_DESCRIPTION = `Shows the person a job's progress as a percentage and an estimated finish time, read live from the output it writes, so they can see when a long or background job will likely be done. The estimate keeps updating by itself while the job runs; you don't need to check on it.
36
37Call it in the same response as your other tool calls, right after starting a job that runs for minutes and prints countable progress: a background command (pass its task id), or anything else that writes a log (pass a short name and the file). Say how progress shows:
38- done_pattern and total: every output line the pattern matches is one unit of total (for example "^  \\S+ (RESOLVED|no) " with total 90).
39- or progress_pattern: a line holding the progress, with groups (done)(total) or one group with a percentage.
40Common progress bars (tqdm, pytest, cargo, "[3/10]", "step 3 of 10") are read without a pattern. Add expected_minutes when you can estimate the whole job's length; for a job already running, started_minutes_ago.`
41
42let ticker: Timer | undefined
43let busy = false
44let statusShown = false
45let frames: Timer | undefined
46let frameMs = 0
47
48/**
49 * Redraws for the band while it moves: the spinner turns four times a second while a job runs,
50 * and a finished job's glint runs at 80 ms for its one second. Nothing runs when nothing moves.
51 */
52function syncFrames($: EngineInterface, list: readonly Job[], now: number) {
53  const isBeat = list.some(j => j.status === 'completed' && j.endedAt !== undefined && now - j.endedAt < BEAT_MS)
54  const want = isBeat ? 80 : list.some(j => j.status === 'running') ? 250 : 0
55  if (want === frameMs) return
56  frames?.cancel()
57  frames = undefined
58  frameMs = want
59  if (want > 0) frames = $.clock.every(want, () => void frame($).catch(() => undefined))
60}
61
62async function frame($: EngineInterface) {
63  $.ui.invalidate('ui.render')
64  syncFrames($, await read($, jobs), await $.clock.now())
65}
66
67/** What stays: running jobs, a finished or stopped one through its beat, a failure until seen or ten minutes old. */
68function remaining(list: readonly Job[], now: number, seenIds: readonly string[]) {
69  return list.filter(j => {
70    if (j.status === 'running') return true
71    const age = now - (j.endedAt ?? now)
72    if (j.status === 'failed') return !seenIds.includes(j.id) && age < KEEP_FAILED_MS
73    return age < BEAT_MS + 1_000
74  })
75}
76
77/** Reads what the job's file gained since the last look, in chunks, and feeds it to the estimate. */
78async function poll($: EngineInterface, job: Job, now: number): Promise<Job> {
79  if (!job.file) return job
80  const st = await $.fs.stat(job.file).catch(() => undefined)
81  if (!st || st.kind !== 'file') return job
82  let j = st.size < job.offset ? { ...job, offset: 0, carry: '' } : job
83  for (let n = 0; n < CHUNKS_PER_TICK && j.offset < st.size; n++) {
84    const want = Math.min(CHUNK, st.size - j.offset)
85    const r = await $.process.run(['/bin/sh', '-c', 'tail -c +"$1" "$2" | head -c "$3"', 'sh', String(j.offset + 1), job.file, String(want)])
86    if (r.exitCode !== 0) break
87    j = { ...feed(j, r.stdout, now), offset: j.offset + want, grewAt: now }
88    if (j.status !== 'running') break
89  }
90  const measured = j.statusLine || j.frac !== undefined || j.done > 0
91  if (j.status === 'running' && measured && now - (j.grewAt ?? j.startedAt) >= STALE_MS) {
92    j = { ...j, status: 'failed', lastLine: 'no update for 20 min', endedAt: now }
93  }
94  return j
95}
96
97async function tick($: EngineInterface) {
98  if (busy) return
99  busy = true
100  try {
101    const now = await $.clock.now()
102    const polled = new Map<string, Job>()
103    for (const j of await read($, jobs)) {
104      if (j.status === 'running') polled.set(j.id, await poll($, j, now))
105    }
106    const seenIds = await read($, seen)
107    const list = await update($, jobs, cur => remaining(cur.map(c => (c.status === 'running' && polled.get(c.id)) || c), now, seenIds))
108    await showStatus($, list, now)
109    syncFrames($, list, now)
110    $.ui.invalidate('ui.render')
111    if (list.length === 0) {
112      ticker?.cancel()
113      ticker = undefined
114    }
115  } finally {
116    busy = false
117  }
118}
119
120function ensureTicking($: EngineInterface) {
121  if (ticker === undefined) ticker = $.clock.every(POLL_MS, () => void tick($).catch(() => undefined))
122}
123
124/**
125 * The band above the prompt is drawn on the terminal and the desktop app only; mobile and VS Code get
126 * the same news as this plugin's status line, while one of them is attached. A terminal session seen
127 * over Remote Control has neither: there only the tool's reply reaches the viewer.
128 */
129async function showStatus($: EngineInterface, list: readonly Job[], now: number) {
130  const remote = (await $.session.surfaces().catch(() => [] as string[])).some(s => REMOTE.has(s))
131  const running = list.filter(j => j.status === 'running')
132  if (!remote || running.length === 0) {
133    if (statusShown) $.ui.status(undefined)
134    statusShown = false
135    return
136  }
137  const tz = await read($, tzOffset)
138  const parts = running.slice(0, 2).map(j => {
139    const d = describe(j, now, tz)
140    return `${d.label} ${d.progress.replace(/^[▰▱]+ /, '')} ${d.eta}`.replace(/\s+/g, ' ').trim()
141  })
142  const more = running.length > 2 ? ` · +${running.length - 2} more` : ''
143  $.ui.status(`⏳ ${parts.join(' · ')}${more}`)
144  statusShown = true
145}
146
147async function addJob($: EngineInterface, job: Job) {
148  await update($, jobs, cur => [...cur.filter(c => c.id !== job.id), job])
149  ensureTicking($)
150}
151
152async function finish($: EngineInterface, id: string, status: Job['status']) {
153  const now = await $.clock.now()
154  const list = await read($, jobs)
155  const job = list.find(j => j.id === id && j.status === 'running')
156  if (!job) return
157  const last = await poll($, job, now)
158  await update($, jobs, cur => cur.map(c => (c.id === id ? { ...last, status: last.status === 'running' ? status : last.status, endedAt: now } : c)))
159  void tick($).catch(() => undefined)
160}
161
162type TrackInput = {
163  task?: unknown
164  file?: unknown
165  label?: unknown
166  done_pattern?: unknown
167  progress_pattern?: unknown
168  finished_pattern?: unknown
169  total?: unknown
170  expected_minutes?: unknown
171  started_minutes_ago?: unknown
172  stop?: unknown
173  status_line?: unknown
174}
175
176const str = (v: unknown) => (typeof v === 'string' && v.trim() !== '' ? v : undefined)
177
178/** The running background task that writes `file`: its own output file, or a file its command names. */
179function writerOf(list: readonly Job[], file: string): Job | undefined {
180  const name = file.split('/').pop() ?? ''
181  const named = name.length > 3 && new RegExp(`(^|[\\s>=/'"])${name.replace(/[.*+?^${}()|[\]\\]/g, '\\$&')}($|[\\s'";|&)])`)
182  return list.find(j => j.status === 'running' && j.source === 'background'
183    && (j.file === file || (j.command !== undefined && (j.command.includes(file) || (named !== false && named.test(j.command))))))
184}
185const num = (v: unknown) => (typeof v === 'number' && Number.isFinite(v) && v > 0 ? v : undefined)
186
187async function track($: EngineInterface, i: TrackInput): Promise<string> {
188  const id = str(i.task)
189  if (!id) return 'track_progress needs `task`: a background task id, or a short name for the job.'
190  const now = await $.clock.now()
191  const list = await read($, jobs)
192  const named = list.find(j => j.id === id)
193  if (i.stop === true) {
194    await update($, jobs, cur => cur.filter(c => c.id !== id))
195    void tick($).catch(() => undefined)
196    return `Stopped showing ${id}.`
197  }
198  for (const key of ['done_pattern', 'progress_pattern', 'finished_pattern'] as const) {
199    if (str(i[key]) && !compile(str(i[key]))) return `${key} is not a valid regular expression.`
200  }
201  const file = str(i.file) ?? named?.file
202  if (!file) return `No file to read for ${id}: pass the log or output file the job writes, or the id of a background task Claude Code started.`
203  const st = await $.fs.stat(file).catch(() => undefined)
204  if (!st || st.kind !== 'file') return `${file} is not a file I can read.`
205  // A log a running background command writes, tracked under a name of its own: the same job, so it
206  // stays one row, under the task's id (its notification ends it).
207  const existing = named ?? writerOf(list, file)
208  const ago = num(i.started_minutes_ago)
209  const startedAt = existing?.startedAt ?? (ago ? now - ago * 60_000 : now)
210  const fromStart = existing?.source === 'background' || ago !== undefined
211  let job: Job = {
212    id: existing?.id ?? id,
213    label: str(i.label) ?? existing?.label ?? id,
214    source: existing?.source ?? 'claude',
215    file,
216    command: existing?.command,
217    donePattern: str(i.done_pattern) ?? existing?.donePattern,
218    progressPattern: str(i.progress_pattern) ?? existing?.progressPattern,
219    finishedPattern: str(i.finished_pattern) ?? existing?.finishedPattern,
220    statusLine: typeof i.status_line === 'boolean' ? i.status_line : existing?.statusLine,
221    lastLine: existing?.lastLine,
222    total: num(i.total) ?? existing?.total,
223    expectedMs: num(i.expected_minutes) ? (i.expected_minutes as number) * 60_000 : existing?.expectedMs,
224    startedAt,
225    offset: 0,
226    grewAt: now,
227    carry: '',
228    done: 0,
229    samples: fromStart ? [[startedAt, 0]] : [],
230    status: 'running',
231  }
232  job = await poll($, job, now)
233  await addJob($, job)
234  const tz = await read($, tzOffset)
235  const d = describe(job, now, tz)
236  const read_ = d.progress ? `${d.progress.replace(/^[▰▱]+ /, '')}, ${d.eta}` : d.eta
237  const as = job.id === id ? '' : ` (as background task ${job.id}, which writes it)`
238  return `Showing "${job.label}" to the person${as}: ${read_}. It updates by itself from ${file}.`
239}
240
241export const register: Register = (on, options) => {
242  if (options.enabled === false) {
243    // Off: a tool registered earlier may outlive the reload, so only its call is answered.
244    on('tool.call', { tool: 'mcp__eta__track_progress' }, () => ({ result: 'eta is turned off (/mods turns it on).' }))
245    return
246  }
247  on('session.start', async ($, e, next) => {
248    await $.tool.register({
249      name: TOOL,
250      description: TOOL_DESCRIPTION,
251      isDeferred: false,
252      inputSchema: {
253        type: 'object',
254        properties: {
255          task: { type: 'string', description: 'The background task id, or a short name for a job started another way.' },
256          file: { type: 'string', description: 'The log or output file the job writes; not needed for a background task Claude Code started.' },
257          label: { type: 'string', description: 'A few words the person will recognise the job by.' },
258          done_pattern: { type: 'string', description: 'Regular expression: each output line it matches is one unit done.' },
259          total: { type: 'number', description: 'How many units make the whole job (with done_pattern).' },
260          progress_pattern: { type: 'string', description: 'Regular expression for a progress line: groups (done)(total), or one group with a percentage.' },
261          finished_pattern: { type: 'string', description: 'Regular expression for a line that means the job finished.' },
262          expected_minutes: { type: 'number', description: 'Your estimate of the whole job’s duration, shown until progress is measurable.' },
263          started_minutes_ago: { type: 'number', description: 'For a job already running: how long ago it started.' },
264          status_line: { type: 'boolean', description: "Show the file's last line beside the job: for jobs that write their state rather than countable progress." },
265          stop: { type: 'boolean', description: 'Stop showing this job.' },
266        },
267        required: ['task'],
268      },
269    })
270    const tz = await $.process.run(['/bin/date', '+%z']).catch(() => undefined)
271    const m = tz?.stdout.trim().match(/^([+-])(\d{2})(\d{2})$/)
272    if (m) await update($, tzOffset, () => (m[1] === '-' ? -1 : 1) * (Number(m[2]) * 60 + Number(m[3])))
273    if ((await read($, jobs)).length > 0) ensureTicking($)
274    return next(e)
275  })
276
277  on('tool.call', async ($, e, next) => {
278    // A subagent's tool call carries its agentId; the main loop's does not.
279    const fromSubagent = (e as { agentId?: unknown }).agentId !== undefined
280    if (e.tool.endsWith(`__${TOOL}`)) {
281      if (fromSubagent) return { result: 'track_progress shows jobs to the person from the main conversation only; nothing is shown for a subagent\'s job.' }
282      return { result: await track($, e as TrackInput) }
283    }
284    const r = await next(e)
285    if (r.deny !== undefined) return r
286    if (e.tool === 'TaskStop' && !r.isError) {
287      const input = e as { task_id?: unknown; shell_id?: unknown }
288      const id = str(input.task_id) ?? str(input.shell_id)
289      if (id) await finish($, id, 'killed')
290      return r
291    }
292    if (e.tool !== 'Bash' || r.isError) return r
293    const id = (r.result as { backgroundTaskId?: unknown } | undefined)?.backgroundTaskId
294    if (typeof id !== 'string' || fromSubagent) return r
295    const input = e as { command?: unknown; description?: unknown }
296    const now = await $.clock.now()
297    const file = (r.text ?? '').match(/written to:?\s+(\S+?\.output)\b/)?.[1]
298    const label = str(input.description) ?? String(input.command ?? id).split('\n')[0] ?? id
299    await addJob($, {
300      id, label, source: 'background', file, command: str(input.command), startedAt: now, offset: 0, carry: '', done: 0,
301      samples: [[now, 0]], status: 'running',
302    })
303    const hint = `Background task ${id} now shows the person its elapsed time. If its output will show countable progress, call ${TOOL} for it in your next response, alongside your other calls, so they also see a percentage and an ETA.`
304    return { ...r, context: [...(r.context ?? []), hint] }
305  }).catch(($, e, next) => next(e))
306
307  on('prompt.submit', async ($, e, next) => {
308    // The person writes again: the failures on the band have been seen, and leave it.
309    if (e.origin.kind === 'composer' || e.origin.kind === 'bridge') {
310      const failed = (await read($, jobs)).filter(j => j.status === 'failed').map(j => j.id)
311      if (failed.length > 0) await update($, seen, s => [...new Set([...s, ...failed])])
312    }
313    if (e.origin.kind === 'task-notification') {
314      for (const j of await read($, jobs)) {
315        if (j.status !== 'running' || j.source !== 'background' || !e.text.includes(j.id)) continue
316        const word = e.text.match(/<status>\s*(\w+)/)?.[1] ?? 'completed'
317        await finish($, j.id, word === 'failed' ? 'failed' : word === 'killed' || word === 'stopped' ? 'killed' : 'completed')
318      }
319    }
320    return next(e)
321  }).catch(($, e, next) => next(e))
322
323  // Its own tool call reads as one line: what is shown, from which file. Where the band carries the
324  // progress (terminal, desktop), the tool's reply draws nothing under it; elsewhere it is the only news.
325  on('ui.render', { component: 'ToolUse', props: { tool: `mcp__eta__${TOOL}` } }, async ($, e, next) => {
326    const i = (e.props.input ?? {}) as TrackInput
327    const task = str(i.task)
328    if (!task) return next(e)
329    const { Box, Text } = $.ui.resolve(e)
330    const file = str(i.file)?.split('/').pop()
331    const total = num(i.total)
332    const what = i.stop === true ? `stopped showing ${task}` : `showing ${str(i.label) ?? task} above the prompt`
333    const detail = i.stop === true ? '' : [total ? `${total} steps` : '', file ? `from ${file}` : ''].filter(Boolean).join(' · ')
334    return (
335      <Box flexDirection="row" gap={1}>
336        <Text color={COLORS.accent}>⏳</Text>
337        <Text>{what}</Text>
338        {detail ? <Text color={COLORS.faint}>{`· ${detail}`}</Text> : null}
339      </Box>
340    )
341  }).catch(($, e, next) => next(e))
342
343  on('ui.render', { component: 'ToolResult', props: { tool: `mcp__eta__${TOOL}` } }, async ($, e, next) => {
344    if (e.props.isErrored || (e.surface !== 'terminal' && e.surface !== 'desktop')) return next(e)
345    const text = typeof e.props.output === 'string' ? e.props.output : ''
346    // A reply that is not the usual "Showing … to the person" (a bad pattern, no file) stays readable.
347    if (text !== '' && !/^(Showing|Stopped showing)/.test(text)) return next(e)
348    const { Box } = $.ui.resolve(e)
349    return <Box />
350  }).catch(($, e, next) => next(e))
351
352  on('ui.render', { component: 'AbovePrompt' }, async ($, e, next) => {
353    const list = await read($, jobs)
354    if (e.props.hasSurvey || list.length === 0) return next(e)
355    const now = await $.clock.now()
356    const view = { seen: await read($, seen) }
357    const rows = table(list, now, await read($, tzOffset), e.props.bodyColumns, view)
358    if (rows.length === 0) return next(e)
359    syncFrames($, list, now)
360    const { Box, Text } = $.ui.resolve(e)
361    const drawn = rows.map(r => (
362      <Box key={r.key} flexDirection="row">
363        {r.parts.map(p => (
364          <Text color={COLORS[p.tone]} bold={p.tone === 'glint'}>
365            {p.text}
366          </Text>
367        ))}
368      </Box>
369    ))
370    const below = await next(e)
371    return (
372      <Box flexDirection="column">
373        {drawn}
374        {below ?? null}
375      </Box>
376    )
377  }).catch(($, e, next) => next(e))
378}
379
hooks/estimate.ts 181 lines
1import type { Job } from '../types'
2
3export type { Job }
4
5/** Progress lines common tools print, tried when Claude named no pattern. */
6const AUTO: { re: RegExp; read: (m: RegExpMatchArray) => number | undefined }[] = [
7  { re: /(\d{1,3}(?:\.\d+)?)%\|/, read: m => pct(m[1]) },                                  // tqdm
8  { re: /\[\s*(\d{1,3})%\]/, read: m => pct(m[1]) },                                        // pytest
9  { re: /Building \[[^\]]*\]\s*(\d+)\/(\d+)/, read: m => ratio(m[1], m[2]) },              // cargo
10  { re: /\[(\d+)\/(\d+)\]/, read: m => ratio(m[1], m[2]) },                                 // ninja, [3/10]
11  { re: /\b(?:step|task|test|file|item|batch|epoch|attempt|block|run|job|shard|chunk|page)s?\s*:?\s*(\d+)\s*(?:\/|of)\s*(\d+)/i, read: m => ratio(m[1], m[2]) },
12  { re: /\b(?:progress|done|completed?)\b\D{0,12}(\d{1,3}(?:\.\d+)?)\s*%/i, read: m => pct(m[1]) },
13]
14
15function pct(s: string | undefined) {
16  const v = Number(s)
17  return Number.isFinite(v) && v >= 0 && v <= 100 ? v / 100 : undefined
18}
19
20function ratio(a: string | undefined, b: string | undefined) {
21  const x = Number(a)
22  const y = Number(b)
23  return y > 0 && x >= 0 && x <= y ? x / y : undefined
24}
25
26/** Reads one progress line with a pattern Claude gave: named or positional groups. */
27function readPattern(m: RegExpMatchArray) {
28  const g = m.groups ?? {}
29  if (g.done !== undefined && g.total !== undefined) return ratio(g.done, g.total)
30  if (g.pct !== undefined) return pct(g.pct)
31  if (m[2] !== undefined) return ratio(m[1], m[2])
32  return pct(m[1])
33}
34
35export function compile(source: string | undefined) {
36  if (!source) return undefined
37  try {
38    return new RegExp(source)
39  } catch {
40    return undefined
41  }
42}
43
44/**
45 * Feeds new output to a job: counts done lines, reads the latest progress line, notices a finish
46 * line. A partial last line waits in `carry` for the rest. Returns the updated job.
47 */
48export function feed(job: Job, text: string, now: number): Job {
49  const all = job.carry + text
50  const parts = all.split(/\r\n|\r|\n/)
51  const carry = /[\r\n]$/.test(all) ? '' : (parts.pop() ?? '')
52  if (carry === '' && parts.at(-1) === '') parts.pop()
53  const done = compile(job.donePattern)
54  const progress = compile(job.progressPattern)
55  const finished = compile(job.finishedPattern)
56  let count = job.done
57  let frac = job.frac
58  let ended = false
59  let lastLine = job.lastLine
60  // A bar redrawn with \r is still mid-line: read it too, without keeping it.
61  for (const line of [...parts, carry]) {
62    if (line.trim() !== '') lastLine = line.trim()
63    if (line !== carry && done?.test(line)) count++
64    if (progress) {
65      const m = line.match(progress)
66      const v = m ? readPattern(m) : undefined
67      if (v !== undefined) frac = v
68    } else if (!done) {
69      for (const { re, read } of AUTO) {
70        const m = line.match(re)
71        const v = m ? read(m) : undefined
72        if (v !== undefined) {
73          frac = v
74          break
75        }
76      }
77    }
78    if (line !== carry && finished?.test(line)) ended = true
79  }
80  if (done && job.total) frac = Math.min(1, count / job.total)
81  const moved = frac !== undefined && frac !== job.frac
82  const samples = moved ? thin([...job.samples, [now, frac as number]]) : job.samples
83  let next: Job = { ...job, carry: carry.slice(-4000), lastLine, done: count, frac, samples, lastProgressAt: moved ? now : job.lastProgressAt }
84  if (ended || (frac !== undefined && frac >= 1)) next = { ...next, status: 'completed', endedAt: now }
85  return next
86}
87
88/** Keeps the first sample and at most 120 after it. */
89function thin(samples: [number, number][]) {
90  return samples.length <= 121 ? samples : [samples[0] as [number, number], ...samples.slice(-120)]
91}
92
93export type Estimate = {
94  frac?: number
95  /** When it will likely finish, and a range from the overall and the recent rate. */
96  at?: number
97  lo?: number
98  hi?: number
99  basis: 'measured' | 'guess' | 'none'
100  /** How long since progress last moved, once that is unusually long. */
101  stalledMs?: number
102}
103
104export function estimate(job: Job, now: number): Estimate {
105  const frac = job.frac
106  const stalledMs = stalled(job, now)
107  const first = job.samples[0]
108  if (frac !== undefined && first && frac > first[1] && now > first[0]) {
109    const elapsed = now - first[0]
110    const overall = (frac - first[1]) / elapsed
111    const window = Math.max(elapsed / 4, 60_000)
112    const back = [...job.samples].reverse().find(([t]) => t <= now - window) ?? first
113    const recent = (frac - back[1]) / (now - back[0])
114    const remaining = 1 - frac
115    let mid = remaining / overall
116    if (job.expectedMs && frac < 0.1) {
117      const w = frac / 0.1
118      mid = w * mid + (1 - w) * Math.max(0, job.startedAt + job.expectedMs - now)
119    }
120    const fast = Math.max(overall, recent)
121    const slow = Math.min(overall, recent)
122    return {
123      frac, at: now + mid, lo: now + Math.min(mid, remaining / fast),
124      hi: slow > 0 ? now + Math.max(mid, remaining / slow) : undefined, basis: 'measured', stalledMs,
125    }
126  }
127  if (job.expectedMs) return { frac, at: job.startedAt + job.expectedMs, basis: 'guess', stalledMs }
128  return { frac, basis: 'none', stalledMs }
129}
130
131/** Progress that has not moved for three times its usual gap, and at least two minutes. */
132function stalled(job: Job, now: number) {
133  if (job.status !== 'running' || job.lastProgressAt === undefined || job.samples.length < 3) return undefined
134  const first = job.samples[0] as [number, number]
135  const gap = (job.lastProgressAt - first[0]) / (job.samples.length - 1)
136  const quiet = now - job.lastProgressAt
137  return quiet > Math.max(3 * gap, 120_000) ? quiet : undefined
138}
139
140export function duration(ms: number) {
141  const m = Math.max(1, Math.round(ms / 60_000))
142  if (m < 60) return `${m}m`
143  return `${Math.floor(m / 60)}h${String(m % 60).padStart(2, '0')}`
144}
145
146/** hh:mm in the person's time zone, given its offset from UTC in minutes. */
147export function clock(ms: number, tzOffsetMin: number) {
148  const d = new Date(ms + tzOffsetMin * 60_000)
149  return `${String(d.getUTCHours()).padStart(2, '0')}:${String(d.getUTCMinutes()).padStart(2, '0')}`
150}
151
152export function bar(frac: number, cells = 10) {
153  const full = Math.max(0, Math.min(cells, Math.round(frac * cells)))
154  return '▰'.repeat(full) + '▱'.repeat(cells - full)
155}
156
157/** One line for a job: what a status line or a band row says. */
158export function describe(job: Job, now: number, tz: number | null) {
159  const label = job.label.length > 32 ? job.label.slice(0, 31) + '…' : job.label
160  if (job.status !== 'running') {
161    const took = duration((job.endedAt ?? now) - job.startedAt)
162    const word = job.status === 'completed' ? `done in ${took}` : job.status === 'failed' ? `failed after ${took}` : `stopped after ${took}`
163    return { label, progress: '', eta: word }
164  }
165  const e = estimate(job, now)
166  const count = job.donePattern && job.total ? ` ${job.done}/${job.total}` : ''
167  const progress = e.frac === undefined ? '' : `${bar(e.frac)} ${Math.floor(e.frac * 100)}%${count}`
168  const at = (ms: number) => (tz === null ? `in ${duration(ms - now)}` : clock(ms, tz))
169  let eta: string
170  if (e.basis === 'measured' && e.at !== undefined) {
171    const range = e.lo !== undefined && e.hi !== undefined && e.hi - e.lo >= 4 * 60_000 ? ` (${at(e.lo)}–${at(e.hi)})` : ''
172    eta = `ETA ${at(e.at)}${tz === null ? '' : ` · in ${duration(e.at - now)}`}${range}`
173  } else if (e.basis === 'guess' && e.at !== undefined) {
174    eta = e.at > now ? `ETA ~${at(e.at)} (Claude's guess)` : `past Claude's guess of ${duration((job.expectedMs ?? 0))}`
175  } else {
176    eta = `running ${duration(now - job.startedAt)}${e.frac === undefined ? ' · no progress readable yet' : ' · measuring'}`
177  }
178  if (e.stalledMs !== undefined) eta += ` · no progress for ${duration(e.stalledMs)}`
179  return { label, progress, eta }
180}
181
hooks/table.ts 265 lines
1// The band's table of jobs as data: rows of toned parts, in columns sized to the band's width.
2// Drawing them is register.tsx's job.
3//
4// Columns, left to right: state, name, gauge, percent, done/total, ETA clock, "in …", range, note.
5// Running jobs come first, soonest ETA first. What has run is done with: a finished job shows its
6// full gauge for a beat and leaves; a failed one stays in red below a hairline until the person
7// has seen it. Columns keep their widths while the values in them change; when the band narrows,
8// columns go in a fixed order.
9//
10// One palette with smart-compact's line: the Claude accent for what runs, green for done, red for
11// failed, the readable grey for labels and the dim one for what sits back.
12
13import type { Job } from '../types'
14import { clock, duration, estimate } from './estimate'
15
16export type Tone = 'accent' | 'good' | 'bad' | 'warn' | 'text' | 'label' | 'faint' | 'glint'
17
18export type Part = { text: string; tone: Tone }
19
20/** A job row, the hairline, or the line that folds the ended jobs (a button). */
21export type Row = { key: string; kind: 'job' | 'rule' | 'fold'; parts: Part[] }
22
23/** How long a finished job's gauge shows full, with one glint along it, before it becomes a done row. */
24export const BEAT_MS = 1200
25const SPIN = ['◐', '◓', '◑', '◒'] as const
26const GAP = '  '
27
28function part(text: string, tone: Tone): Part {
29  return { text, tone }
30}
31
32export function width(text: string): number {
33  return [...text].length
34}
35
36/** Cut to `n` cells with an ellipsis, or padded to `n`. */
37export function cell(text: string, n: number): string {
38  const chars = [...text]
39  if (chars.length <= n) return text + ' '.repeat(n - chars.length)
40  return chars.slice(0, Math.max(0, n - 1)).join('') + '…'
41}
42
43function right(text: string, n: number): string {
44  return ' '.repeat(Math.max(0, n - width(text))) + text
45}
46
47/** smart-compact's gauge: half-cell steps (━ full, ╸ half, ─ empty), with an optional glint cell. */
48export function gauge(frac: number, cells: number, tone: Tone, glintAt?: number): Part[] {
49  const halves = Math.round(Math.min(1, Math.max(0, frac)) * cells * 2)
50  const out: Part[] = []
51  for (let i = 0; i < cells; i += 1) {
52    const filled = Math.min(2, Math.max(0, halves - i * 2))
53    const glyph = filled === 2 ? '━' : filled === 1 ? '╸' : '─'
54    const isGlint = glintAt !== undefined && Math.abs(i - glintAt) < 1
55    const t: Tone = isGlint ? (filled > 0 ? 'glint' : 'label') : filled > 0 ? tone : 'faint'
56    const last = out.at(-1)
57    if (last && last.tone === t) last.text += glyph
58    else out.push(part(glyph, t))
59  }
60  return out
61}
62
63type State = 'running' | 'waiting' | 'beat' | 'completed' | 'failed' | 'killed'
64
65/** What one job's row says, column by column. */
66export type Info = {
67  job: Job
68  state: State
69  frac?: number
70  pct?: string
71  count?: string
72  at?: string
73  inn?: string
74  innTone?: Tone
75  range?: string
76  note?: string
77  noteTone?: Tone
78  /** The note is the job's own last line: cut to the room there is. */
79  liveLine?: boolean
80  /** When it will likely finish, for the order. */
81  eta?: number
82}
83
84export function info(job: Job, now: number, tz: number | null): Info {
85  const count = job.donePattern && job.total ? `${job.done}/${job.total}` : undefined
86  if (job.status !== 'running') {
87    const end = job.endedAt ?? now
88    if (job.status === 'completed' && now - end < BEAT_MS) return { job, state: 'beat', frac: 1, pct: '100%', count }
89    const word = job.status === 'completed' ? 'done' : job.status === 'failed' ? 'failed' : 'stopped'
90    const took = duration(end - job.startedAt)
91    return {
92      job,
93      state: job.status,
94      at: tz === null ? word : `${word} ${clock(end, tz)}`,
95      inn: job.status === 'completed' ? `took ${took}` : `after ${took}`,
96    }
97  }
98  const e = estimate(job, now)
99  const pct = e.frac === undefined ? undefined : `${Math.floor(e.frac * 100)}%`
100  const stalledNote = e.stalledMs === undefined ? {} : { note: `no progress for ${duration(e.stalledMs)}`, noteTone: 'warn' as const }
101  if (e.basis === 'measured' && e.at !== undefined) {
102    const wide = e.lo !== undefined && e.hi !== undefined && e.hi - e.lo >= 4 * 60_000 && tz !== null
103    return {
104      job, state: 'running', frac: e.frac, pct, count, eta: e.at,
105      at: tz === null ? undefined : clock(e.at, tz),
106      inn: `in ${duration(e.at - now)}`,
107      range: wide ? `${clock(e.lo as number, tz as number)}–${clock(e.hi as number, tz as number)}` : undefined,
108      ...stalledNote,
109    }
110  }
111  if (e.basis === 'guess' && e.at !== undefined) {
112    const isOver = e.at <= now
113    return {
114      job, state: 'running', frac: e.frac, pct, count, eta: e.at,
115      at: tz === null || isOver ? undefined : `~${clock(e.at, tz)}`,
116      inn: isOver ? `over by ${duration(now - e.at)}` : `in ~${duration(e.at - now)}`,
117      innTone: isOver ? 'warn' : undefined,
118      note: "Claude's guess",
119      noteTone: 'faint',
120      ...stalledNote,
121    }
122  }
123  const line = job.statusLine && e.frac === undefined && job.lastLine ? job.lastLine.trim() : undefined
124  return {
125    job, state: e.frac === undefined ? 'waiting' : 'running', frac: e.frac, pct, count,
126    inn: `${duration(now - job.startedAt)} so far`,
127    note: line ?? (e.frac === undefined ? 'no progress readable yet' : 'measuring'),
128    noteTone: line ? 'text' : 'faint',
129    liveLine: line !== undefined,
130    ...stalledNote,
131  }
132}
133
134/** The columns a table keeps at a width. */
135export type Layout = { name: number; cells: number; count: boolean; inn: boolean; range: boolean; note: boolean }
136
137type Widths = { count: number; at: number; inn: number; range: number; note: number }
138
139function rowWidth(l: Layout, w: Widths): number {
140  return (
141    3 + l.name + 2 + l.cells + 2 + 4 +
142    (l.count && w.count ? 2 + w.count : 0) +
143    (w.at ? 2 + w.at : 0) +
144    (l.inn && w.inn ? 2 + w.inn : 0) +
145    (l.range && w.range ? 2 + w.range : 0) +
146    (l.note && w.note ? 2 + w.note : 0)
147  )
148}
149
150/**
151 * The widest layout that fits: the note goes first, then the range, the gauge shrinks to 12,
152 * done/total goes, then "in …", the gauge shrinks to 6, and last the name narrows.
153 */
154export function layout(longestName: number, w: Widths, columns: number): Layout {
155  let l: Layout = { name: Math.min(28, Math.max(8, longestName)), cells: 20, count: true, inn: true, range: true, note: true }
156  const steps: ((l: Layout) => Layout)[] = [
157    l => ({ ...l, note: false }),
158    l => ({ ...l, range: false }),
159    l => ({ ...l, cells: 12 }),
160    l => ({ ...l, count: false }),
161    l => ({ ...l, inn: false }),
162    l => ({ ...l, cells: 6 }),
163    l => ({ ...l, name: Math.min(l.name, 16) }),
164    l => ({ ...l, name: Math.min(l.name, 10) }),
165  ]
166  for (const step of steps) {
167    if (rowWidth(l, w) <= columns) break
168    l = step(l)
169  }
170  return l
171}
172
173/** The failures the person has seen (written again since): those leave the band too. */
174export type View = { seen: readonly string[] }
175
176function iconOf(i: Info, now: number): Part {
177  if (i.state === 'running' || i.state === 'waiting') return part(SPIN[Math.floor(now / 250) % SPIN.length] ?? '◐', 'accent')
178  if (i.state === 'beat' || i.state === 'completed') return part('✓', 'good')
179  if (i.state === 'failed') return part('✕', 'bad')
180  return part('⊘', 'label')
181}
182
183/**
184 * The table, row by row: running jobs (soonest first), then a hairline and the failures the person
185 * has not seen yet. A job that finished or was stopped is gone after its beat.
186 */
187export function table(list: readonly Job[], now: number, tz: number | null, columns: number, view: View): Row[] {
188  const infos = list.map(j => info(j, now, tz))
189  const running = infos
190    .filter(i => i.state === 'running' || i.state === 'waiting' || i.state === 'beat')
191    .sort((a, b) => (a.state === 'beat' ? -1 : b.state === 'beat' ? 1 : (a.eta ?? Infinity) - (b.eta ?? Infinity)))
192  const shown = infos
193    .filter(i => i.state === 'failed' && !view.seen.includes(i.job.id))
194    .sort((a, b) => (b.job.endedAt ?? 0) - (a.job.endedAt ?? 0))
195
196  const max = (f: (i: Info) => string | undefined) => Math.max(0, ...running.map(i => width(f(i) ?? '')))
197  // Widths that hold still while the values change: done/total as wide as its total twice, the
198  // clock with room for its "~", "in …" at least ten cells, the range its full length.
199  const totals = running.map(i => (i.job.donePattern && i.job.total ? width(String(i.job.total)) * 2 + 1 : 0))
200  const w: Widths = {
201    count: Math.max(0, ...totals),
202    at: running.some(i => i.at) ? Math.max(6, max(i => i.at)) : 0,
203    inn: running.some(i => i.inn) ? Math.max(10, max(i => i.inn)) : 0,
204    range: running.some(i => i.range) ? 11 : 0,
205    note: Math.max(0, ...running.map(i => Math.min(width(i.note ?? ''), i.liveLine ? 40 : Infinity))),
206  }
207  const longest = Math.max(0, ...[...running, ...shown].map(i => width(i.job.label)))
208  const l = layout(longest, w, columns)
209  const rows: Row[] = []
210
211  for (const i of running) {
212    const isBeat = i.state === 'beat'
213    const glintAt = isBeat ? ((now - (i.job.endedAt ?? now)) / BEAT_MS) * (l.cells + 4) - 2 : undefined
214    const parts: Part[] = [iconOf(i, now), part(GAP, 'faint'), part(cell(i.job.label, l.name), 'text'), part(GAP, 'faint')]
215    parts.push(...gauge(i.frac ?? 0, l.cells, isBeat ? 'good' : 'accent', glintAt))
216    parts.push(part(GAP + right(i.pct ?? '—', 4), i.pct ? 'text' : 'faint'))
217    if (l.count && w.count) parts.push(part(GAP + right(i.count ?? '', w.count), 'label'))
218    if (w.at) parts.push(part(GAP + right(i.at ?? '', w.at), 'text'))
219    if (l.inn && w.inn) parts.push(part(GAP + right(i.inn ?? '', w.inn), i.innTone ?? 'label'))
220    if (l.range && w.range) parts.push(part(GAP + cell(i.range ?? '', w.range), 'faint'))
221    if (l.note && w.note && i.note) {
222      const room = Math.max(8, columns - rowWidth(l, { ...w, note: 0 }) - GAP.length)
223      parts.push(part(GAP + (i.liveLine ? cell(i.note, Math.min(room, width(i.note))).trimEnd() : i.note), i.noteTone ?? 'faint'))
224    }
225    rows.push({ key: i.job.id, kind: 'job', parts: trimEnd(parts) })
226  }
227
228  if (running.length > 0 && shown.length > 0) {
229    rows.push({ key: 'rule', kind: 'rule', parts: [part('─'.repeat(Math.min(columns, rowWidth(l, w))), 'faint')] })
230  }
231
232  for (const i of shown) {
233    const failed = i.state === 'failed'
234    rows.push({
235      key: i.job.id,
236      kind: 'job',
237      parts: [
238        iconOf(i, now),
239        part(GAP, 'faint'),
240        part(cell(i.job.label, l.name), failed ? 'bad' : 'label'),
241        part(GAP, 'faint'),
242        part(i.at ?? '', failed ? 'bad' : i.state === 'completed' ? 'good' : 'label'),
243        part(` · ${i.inn ?? ''}`, 'faint'),
244      ],
245    })
246  }
247
248  return rows
249}
250
251/** No trailing blanks: a padded empty column at the end of a row is just room. */
252function trimEnd(parts: Part[]): Part[] {
253  const out = [...parts]
254  while (out.length > 0) {
255    const last = out.at(-1) as Part
256    const trimmed = last.text.replace(/\s+$/, '')
257    if (trimmed) {
258      out[out.length - 1] = { ...last, text: trimmed }
259      break
260    }
261    out.pop()
262  }
263  return out
264}
265
types/index.d.ts 47 lines
1/** A job whose progress is read from a file it writes: a background task's output or a log. */
2export type Job = {
3  id: string
4  label: string
5  source: 'background' | 'claude'
6  file?: string
7  /** A background task's command line: a log it writes is that task's progress, not another job. */
8  command?: string
9  /** Each output line this matches counts one unit done. */
10  donePattern?: string
11  /** A progress line: groups (done)(total), or one group holding a percentage. */
12  progressPattern?: string
13  finishedPattern?: string
14  /** Show the file's last line beside the job, for jobs that write their state. */
15  statusLine?: boolean
16  /** The last non-empty line read from the file. */
17  lastLine?: string
18  total?: number
19  /** Claude's own guess at the whole job's duration, used until progress is measurable. */
20  expectedMs?: number
21  startedAt: number
22  offset: number
23  /** When the file last grew (or the job began); a status-line job silent for too long is ended. */
24  grewAt?: number
25  carry: string
26  done: number
27  frac?: number
28  /** [time, fraction done], oldest first; the first may be the job's start at 0. */
29  samples: [number, number][]
30  lastProgressAt?: number
31  status: 'running' | 'completed' | 'failed' | 'killed'
32  endedAt?: number
33}
34
35declare module 'claude-code' {
36  interface PluginState {
37    eta: {
38      /** The jobs shown: running, finishing (their beat), or failed and not yet seen. */
39      jobs: Job[]
40      /** Failed jobs the person has seen, by writing again since; until then a failure keeps its red row. */
41      seen: string[]
42      /** The person's offset from UTC in minutes, for clock times; null until read. */
43      tzOffsetMin: number | null
44    }
45  }
46}
47